gentoo & v7?

Moderators: Site Moderators, FAHC Science Team

Re: gentoo & v7?

Postby keetrainchild » Thu Jan 26, 2012 9:38 pm

I would be glad to try version 7 on Gentoo Linux as well, since that is my main operating system. Does anyone here know how well it works with the G.P.U. folding, especially with Radeon 6000-series cards?
keetrainchild
 
Posts: 1
Joined: Thu Jan 26, 2012 8:06 pm

Re: gentoo & v7?

Postby darkbasic » Thu Jan 26, 2012 10:04 pm

You can't if you don't use the windows client through wine.
darkbasic
 
Posts: 90
Joined: Sun Jan 08, 2012 12:50 pm

Re: gentoo & v7?

Postby bollix47 » Thu Jan 26, 2012 10:30 pm

AFAIK ... there is no way to run folding@home on an ati graphics card in Linux at this time.
bollix47
 
Posts: 2871
Joined: Sun Dec 02, 2007 6:04 am
Location: Canada

Re: gentoo & v7?

Postby bruce » Sat Jan 28, 2012 3:32 am

FAH doesn't officially support gentoo. V7 has two Linux installers, one for Debian / Mint / Ubuntu and the other for RedHat / CentOS / Fedora. There may be knowledgable Linux folks who will figure out how to tweak one of those installers to work on gentoo but since everything is still in beta, the install scripts are probably moving targets.

Even if you do get V7 installed (and reinstalled, as it progresses toward an eventual public release) it will be for CPU only. V7, with the proper drivers, is ready to support GPUs, but there are no Projects being assigned because the GPU FahCore are not available for Linux.

With respect to graphics support in native linux, here's a 5-page topic that you might be interested in:
viewtopic.php?f=67&t=19896.
bruce
 
Posts: 20355
Joined: Thu Nov 29, 2007 11:13 pm
Location: So. Cal.

Re: gentoo & v7?

Postby darkbasic » Sat Jan 28, 2012 12:19 pm

bruce wrote:With respect to graphics support in native linux


What's the problem? Proprietary drivers have a 90% shared code base among different OS, especially the nvidia one. There will probably be some bugs initially, but they will surely be fixed. AMD is actively working developing OpenCL even on the FOSS driver...
darkbasic
 
Posts: 90
Joined: Sun Jan 08, 2012 12:50 pm

Re: gentoo & v7?

Postby PantherX » Sat Jan 28, 2012 1:07 pm

darkbasic wrote:
bruce wrote:With respect to graphics support in native linux


What's the problem?...
The issue here is the limited developer resources that PG Members have. Moreover, the FahCores may sound "simple" in nature but they have to be 100% scientifically accurate on every single system that it can run on. This isn't an easy feat and it takes time. The reason for the scientific reliability is that unlike other DC Projects, all the WUs are processed once (unless there are some special circumstances) and the Project has a serial nature to it. If a FahCore on system 1 produces a scientific error that isn't detected by the FahCore itself, the error would be carried forward in that Project which might render the results invalid.

If you are talking about porting the FahCores from Windows to OSX/Linux, it again isn't simple as you still need to validate the FahCore. Also, the current GPU FahCores are being worked upon so porting a version that can still be improved upon is a waste of developer time. It makes much more sense to port the FahCore once they are fully matured. When that happens, is anyone's guess.
ETA:
Now ↞ Very Soon ↔ Soon ↔ Soon-ish ↔ Not Soon ↠ End Of Time

Welcome To The F@H Support Forum Ӂ Troubleshooting Bad WUs Ӂ Troubleshooting Server Connectivity Issues
User avatar
PantherX
Site Moderator
 
Posts: 6985
Joined: Wed Dec 23, 2009 10:33 am
Location: Land Of The Long White Cloud

Re: gentoo & v7?

Postby darkbasic » Sat Jan 28, 2012 5:32 pm

PantherX wrote:unlike other DC Projects, all the WUs are processed once


By the way, I've always wondered: how do you deal with WUs processed by peoples with unstable overclocks? Other projects simply discard the bad ones because they process the same WU three times...
darkbasic
 
Posts: 90
Joined: Sun Jan 08, 2012 12:50 pm

Re: gentoo & v7?

Postby bruce » Sat Jan 28, 2012 7:36 pm

If every WU was processed three times, FAH would get only 10% - 15% as much work done. First of all, 2/3 of the processing power would be wasted, and then the trajectory could not proceed to the next generation until the slowest of the three completed the assignment.

It's well worth investing in several levels of software that check for unstable overclocks or any number of other issues that might lead to a WU being discarded. Early detection of a possible corruption helps, too, because the sooner that WU can be issued to some other machine that will produce accurate results, the faster that generation can be completed.

Unlike other projects, FAH does everything reasonable to avoid assigning duplicate WUs. Some of those "other projects" have been known to run out of new work and to assign the same WUs more than three times (or whatever their standard policy is). On the other hand, without too much work by the researchers, FAH could probably find work that is both unique and scientifically valuable for maybe ten times as many computers as currently contribute to FAH. That couldn't happen all at once because they'd need 10x the server power. They would need a significant hardware upgrade in at their computer rooms, but if the rate at which new people join FAH increased significantly, I'm sure they'd figure out how to handle it at the sever level reasonably quickly. A server shortage is the kind of problems that FAH would really LIKE to have. (Their long range plan anticipates a regular growth in donor numbers and in average client processing speed, coupled with a periodic expansion of server capacity.)
bruce
 
Posts: 20355
Joined: Thu Nov 29, 2007 11:13 pm
Location: So. Cal.

Re: gentoo & v7?

Postby PantherX » Sun Jan 29, 2012 12:11 pm

darkbasic wrote:
PantherX wrote:unlike other DC Projects, all the WUs are processed once


By the way, I've always wondered: how do you deal with WUs processed by peoples with unstable overclocks? Other projects simply discard the bad ones because they process the same WU three times...
If you want another prespective, then you may want to read this thread -> viewtopic.php?f=19&t=16526
User avatar
PantherX
Site Moderator
 
Posts: 6985
Joined: Wed Dec 23, 2009 10:33 am
Location: Land Of The Long White Cloud

Re: gentoo & v7?

Postby darkbasic » Sun Jan 29, 2012 12:57 pm

Thanks, but you can't control all your users and a wu will not necessarily fail, the client may simply report wrong results: how do you know if something wrong happened and the wu must be reprocessed?
darkbasic
 
Posts: 90
Joined: Sun Jan 08, 2012 12:50 pm

Re: gentoo & v7?

Postby PantherX » Sun Jan 29, 2012 8:25 pm

darkbasic wrote:...how do you know if something wrong happened and the wu must be reprocessed?
Each WU has two sets of deadlines; Preferred and Final (Details). If the WU isn't returned by the Preferred Deadline, then the same WU is reissued. If a WU returns an error code, it is reissued again for each error code to ensure that it isn't unstable hardware. Do note that the number of times a single WU is reassigned can be limited by the researcher to ensure that bad WUs don't "float around" amongst donors.
User avatar
PantherX
Site Moderator
 
Posts: 6985
Joined: Wed Dec 23, 2009 10:33 am
Location: Land Of The Long White Cloud

Re: gentoo & v7?

Postby 7im » Mon Jan 30, 2012 6:49 pm

PantherX wrote:
darkbasic wrote:...how do you know if something wrong happened and the wu must be reprocessed?
Each WU has two sets of deadlines; Preferred and Final (Details). If the WU isn't returned by the Preferred Deadline, then the same WU is reissued. If a WU returns an error code, it is reissued again for each error code to ensure that it isn't unstable hardware. Do note that the number of times a single WU is reassigned can be limited by the researcher to ensure that bad WUs don't "float around" amongst donors.


I think db is asking how Stanford knows that a completed WU, uploaded without error, doesn't have any data corruption... how does stanford error correct for potential data corruption without having to compare results between multiple computers? ;)

While off-topic, it's still good question, answered several times before. I don't have a previous example bookmarked, do you?
How to provide enough information to get helpful support
Tell me and I forget. Teach me and I remember. Involve me and I learn.
User avatar
7im
 
Posts: 10189
Joined: Thu Nov 29, 2007 5:30 pm
Location: Arizona

Re: gentoo & v7?

Postby bruce » Mon Jan 30, 2012 9:18 pm

7im wrote:I think db is asking how Stanford knows that a completed WU, uploaded without error, doesn't have any data corruption... how does stanford error correct for potential data corruption without having to compare results between multiple computers? ;)

While off-topic, it's still good question, answered several times before. I don't have a previous example bookmarked, do you?


Stanford has multiple layers of security. For obvious reasons, they don't publish those details.

They've built a number of validity tests into the FahCore. Just look at the complaints we get about "What caused error XXXX? My machine is completely stable. I've wasted x days."

Assuming an erroneous result isn't captured on the client, they've built other tests into the servers. Just look for complaints about "Server has a problem..."

Assuming an erroneous result gets past the server based tests, the results are eventually digested by some special software that summarizes and analyzes a large batch of individual WU results. Additional validity testing happens at this level, too.

Stanford is confident that the risks of their research being contaminated by inaccurate data are at an acceptably low level. (Remember there are also risks of contamination as using the three-copies method, though at a low enough level, statistically speaking.)

Suppose someone is intent on submitting false results, and suppose they figure out how to break/bypass the client and server levels of security (unlikely, but statistics doesn't let anyone talk about "absolutes"). It's possible that they MIGHT earn illegal points, but if the bad data is detected during the final data analysis, the results still won't be used. Moreover, this may lead to a project to increase security at an earlier level.
bruce
 
Posts: 20355
Joined: Thu Nov 29, 2007 11:13 pm
Location: So. Cal.

Re: gentoo & v7?

Postby darkbasic » Mon Jan 30, 2012 10:33 pm

Thanks, does this apply to the GPU client too? Because I tried to fold a fahcore_16 WU with a *highly* unstable vga overclock and it completed flawlessly, I even get the points...
darkbasic
 
Posts: 90
Joined: Sun Jan 08, 2012 12:50 pm

Re: gentoo & v7?

Postby bruce » Mon Jan 30, 2012 11:27 pm

Yes
bruce
 
Posts: 20355
Joined: Thu Nov 29, 2007 11:13 pm
Location: So. Cal.

PreviousNext

Return to V7.1.52 Windows/Linux

Who is online

Users browsing this forum: No registered users and 2 guests

cron